Fable 5 safety classifier force-downgrades the whole session on defensive-security work (VSCode extension, no reachable feedback route)

Environment: Claude Code VSCode extension · Max subscription · model set to Claude Fable 5.

What happens: During normal coding on a defensive-security / infrastructure-governance codebase, a routine message trips Fable 5's safety classifier. The session is then silently force-downgraded to Opus 4.8 with:

Fable 5's safeguards flagged this message. The safeguards are intentionally broad right now and may flag safe and routine coding, cybersecurity, or biology work.

Why it's a problem:

1. The false-positive category is my whole workload — defensive-security engineering reads as adversarial to a broad filter, so it's constant, not an edge case.
2. Silent mid-session model swap — the paid-for model is pulled from the entire session, not just the flagged message. There's no per-message block option.
3. No reachable feedback route — /feedback isn't available in the VSCode extension, and the support article's "Send feedback" points back to that same missing in-product control.

Requests:

• Higher classifier precision for defensive-security work (this is a false-positive-rate problem, not a capability problem).
• Block the message, not the session — don't remove the selected model from the whole session.
• A verified-user / appeal path for known-legitimate security workloads.
• Make /feedback reachable from the VSCode extension.